Lake level and paleoenvironmental history of Lake Tanganyika, Africa, as inferred from late Holocene and modern stromatolites Andrew S. Cohen* Department of Geosciences, University of Arizona, Tucson, Arizona 85721 Michael R. Talbot Geological Institute, University of Bergen, Allegt. 41, 5007 Bergen, Norway Department of Geological Sciences, Preston Cloud Research Laboratory, Stanley M. Awramik University of California, Santa Barbara, California 93106 David L. Dettman Department of Geosciences, University of Arizona, Tucson, Arizona 85721 Paul Abell Department of Chemistry, University of Rhode Island, Kingston, Rhode Island 02881 ABSTRACT time between the late sixteenth and early nine- history of that lake, and on the late Holocene cli- teenth centuries, lake level fell to perhaps its mate history of Central Africa. Fossil and living stromatolites are abundant lowest level in the past 2800 yr. By the early Lake Tanganyika is located in the western around the margins of Lake Tanganyika, Af- nineteenth century, lake level had begun to rise branch of the African Rift system (Fig. 1). It is the rica, and provide a wealth of paleolimnologic to the overflow level, apparently the result of a largest of the African rift lakes and the second and paleoclimatic information for the late Holo- regional increase in precipitation/evaporation deepest lake in the world. The lake occupies a se- cene. Six lines of evidence show that stromato- ratios. ries of interconnected half-graben basins, the old- lites and cements are precipitating in the lake to- Weak δ18O/δ13C covariance for late Holo- est of which is probably between 9 and 12 Ma day: (1) carbonate saturation state calculations, cene carbonates suggests that the surface eleva- (Cohen et al., 1993). The basinal bedrock lithol- (2) documentation of living stromatolites and tion of the lake has remained near the outlet ogy consists of Proterozoic metasedimentary their depth distribution, (3) new stable isotope level, with only occasional periods of closure. rocks, Karroo (late Paleozoic–early Mesozoic) data showing the lake’s present mixing state However, there is no simple relationship be- nonmarine sedimentary rocks, and late Tertiary and ancient evaporation and inflow balance, tween solute input from Lake Kivu, isotope in- volcanic rocks in the Ruzizi River watershed, oc- (4) new radiocarbon data and a reevaluation of put from Lake Kivu, and lake levels in Lake cupying the north end of the lake basin. apparent 14C ages derived from Lake Tangan- Tanganyika. Lake Kivu waters are the primary The lake is currently an open basin, discharg- yika carbonates, (5) the presence of modern source of major ions in Lake Tanganyika, but ing into the Zaire basin via the Lukuga River. Mg-calcite cements derived from lake waters, are much less important in controlling the δ18O However, most water loss is from evaporation, and (6) the presence of modern, biologically me- and the lake level of Lake Tanganyika. Because and the outlet serves primarily to maintain a rela- diated Mg-calcite precipitates in the lake. the Ruzizi River’s discharge into Lake Tan- tively constant lake level (Hecky and Degens, Lake Tanganyika’s lake levels have been re- ganyika is largely derived from sources other 1973). The lake has been closed as recently as markably stable over the past 2800 yr, fluctuat- than Lake Kivu, the overflow events in the two 1878 (Camus, 1965). ing around the marginally open to marginally lakes have been uncoupled during the late Climate in the lake basin is semihumid tropical. closed level through most of this time period. Holocene. The lake receives continuous, but seasonally vari- Lake lowstands and high δ18O values from the able, inflow from four major rivers (Malagarasi, ninth century B.C. to the early fifth century A.D. INTRODUCTION Ruzizi, Lugufu, and Lufubu), as well as numerous indicate that the lake basin was comparatively small rivers. Most of the influent streams originate dry during this time. However, the period prior Numerous studies have demonstrated that on rift escarpments, close to the lake, and therefore to the most recent opening of Lake Kivu into laminated stromatolites and massive thrombo- carry very low dissolved solid loads. In contrast, the Lake Tanganyika basin (ca. A.D. 550) was lites can be used in paleolimnological recon- the Ruzizi River drains alkaline Lake Kivu to the not marked by major lake lowstands, nor was structions (Abell et al., 1982; Casanova, 1986; north, and is the dominant source for most solutes this opening accompanied by a dramatic lake- Hillaire-Marcel and Casanova, 1987; Casanova in Lake Tanganyika, excluding Ca2+ (Craig et al., level rise. The Kivu opening was roughly coin- and Thouin, 1990; Straccia et al., 1990; Talbot, 1974). Lake Kivu surface waters are enriched in cident with a significant shift toward isotopi- 1990). Stromatolites can provide information on Mg2+ relative to Ca2+, accounting for the unusually cally lighter (δ18O and δ13C) lake water, which lake-level fluctuations (based on their elevation), high Mg/Ca molar ratio (≈9) of the Ruzizi River persists today. The lake remained close to its paleohydrology (through 13C and 18O analyses of where it enters Lake Tanganyika. Ruzizi waters are outlet level between the sixth and thirteenth stromatolitic carbonates), paleoenvironments subsequently diluted by waters from the other in- centuries A.D. Lake levels rose between the (from their growth morphology), and chronology fluent streams to produce a low salinity, low alka- fourteenth and sixteenth centuries. At some (through direct 14C dating). Here we discuss new linity lake water (conductivity = 670 µmho/cm, observations on the stromatolites of Lake Tan- Ac = 6.6 meq/l). Ground-water input from small *E-mail: [email protected] ganyika, Africa, that shed light on the most recent hydrothermal springs has been documented in GSA Bulletin; April 1997; v. 109; no. 4; p. 444–460; 8 figures; 4 tables. 444 PALEOENVIRONMENTAL HISTORY OF LAKE TANGANYIKA some parts of Lake Tanganyika (TANGHYDRO Group, 1992), but the volume of fluid flow is rela- tively small (≈100–300 l/s each, for the two fields discovered to date). Stable isotope (18O and 13C) and major ion chemistry measurements suggest that the surface waters are well mixed (Casanova and Hillaire-Marcel, 1992), although these epilim- netic waters mix much more slowly with the volu- minous and anoxic monimolimnion of the lake (Coulter and Spigel, 1991). Surface water compo- sition and major ion concentrations have been very constant, both seasonally and over several dec- ades, because of the lake’s size and the long resi- dence time of its water (Hecky and Degens, 1973; Cohen and Thouin, 1987; Batungwanayo, 1987; Casanova and Hillaire-Marcel, 1992). Calcareous sediments were first reported from the lake by Dubois (1959). Cores taken in the 1970s showed that aragonitic and calcitic muds have been deposited periodically during the Holo- cene (Stoffers and Hecky, 1978). Cohen and Thouin (1987) described numerous carbonate lithofacies in the lake, including lime muds, coated- grain and skeletal-fragment sands and gravels, and microbially cemented boundstones (stromatolites and thrombolites). Most of these sediments were either high-Mg calcite or aragonite, an observation that is consistent with the carbonate mineralogy predicted by Müller et al. (1972) based on the mo- lal proportions of Mg/Ca in the lake water. Cohen and Thouin (1987) proposed that carbonates are currently being deposited in the lake, and that stro- matolites and thrombolites have formed under es- sentially modern conditions. Subsequent work has shown that many of the surficial carbonate deposits observed on the lake floor are quite old. Cohen (1989) showed that shell lags of up to 2000 yr in age are exposed on the lake floor at littoral-sublittoral depths, apparently the result of winnowing events during lowstands of lake level. On the basis of radiocarbon dating and 18O evidence from stromatolites, Casanova and Thouin (1990) and Casanova and Hillaire-Marcel Figure 1. Location map of Lake Tanganyika showing known stromatolite localities. Much of (1992) argued that all stromatolites found on the the shoreline of Lake Tanganyika is unexplored. Thus, this map should not be construed to in- lake floor (and implicitly all littoral carbonate de- dicate that stromatolites are not present elsewhere. Numbers correspond to locality numbers in posits other than mollusk and ostracode shells) are Table 2. Principal research sites discussed in text are no. 10 (Km 115 site, southern Burundi) and fossil. In their chronology, stromatolites formed no. 14 (Kigoma site, Tanzania). between ca. 2000 B.C. and A.D. 750, during a pe- riod of late Holocene aridity and Lake Kivu clo- sure. They proposed that an increase in the re- precipitated carbonate deposition did not cease in 1987; Casanova and Hillaire-Marcel, 1992). Ac- gional precipitation/evaporation balance and Lake Tanganyika at ca. A.D. 750, but has contin- tivity coefficients were calculated using the subsequent dilution of Lake Tanganyika waters ued to the present. Debye-Hückel equation. Ion pairing in solution during ca. A.D. 650–750 put an end to carbonate was estimated using the thermodynamic associa- accumulation and stromatolite formation. METHODS tion constants preferred by Millero and Schreiber Here we present new data on the late Holocene (1982, their Table 3). We ignored all Na+,K+, and –2 history of Lake Tanganyika. Our findings are pri- Saturation state, water balance, and isotopic SO4 pairs in order to simplify calculations, be- marily based on the occurrence of both modern equilibrium calculations were based on previously cause they had a negligible effect (<1%) on the +2 +2 – –2 and Holocene stromatolites from parts of the lake published water chemistry data (Table 1; Hecky ions of interest (Ca ,Mg , HCO3 and CO3 ). which were not visited by earlier workers. We ar- and Degens, 1973; Craig et al., 1974; Tietze, 1981, By the third iteration of these calculations, changes gue that microbially mediated and inorganically 1982; Cohen and Thouin, 1987; Batungwanayo, in free ion activities were less than 0.04%.